Anaerobic digestion (AD) remains one of the UK’s most promising technologies for generating renewable energy from waste. By using microbial processes to break down organic material, AD produces biogas that can replace fossil fuels and prevent methane emissions from landfills. Yet despite its environmental value, the sector suffers from a fundamental issue: unpredictability.

BiofuelAI was founded to solve this problem. Its origins lie in groundbreaking research conducted at the University of Surrey—research that brought together artificial intelligence, systems biology, and process engineering to tackle AD’s most persistent operational challenge: prediction.

There are over 750 AD sites operating across the UK. Collectively, they contribute around 1% to national greenhouse gas reductions. While meaningful, this figure represents just a fraction of AD’s potential.

The reason is simple but serious: AD systems are inherently complex. Every digester is a live biological reactor, where input materials—from food waste to slurry—interact with diverse microbial populations under constantly shifting conditions. Unlike traditional energy technologies, AD is not easily modelled, and its outputs cannot be controlled with simple setpoints.

Because of this complexity, plant managers tend to minimise risk by purchasing only high-quality, consistent feedstocks, even when these are expensive or seasonal. One poor-quality input can cause days of lost productivity, system instability, or even plant shutdown. This risk aversion leads to high feedstock costs, under-utilised waste streams, and limited sector growth.

In response to these challenges, a research team at the University of Surrey, led by Dr Michael Short, launched a £1.7 million EPSRC/UKRI-funded programme aimed at introducing artificial intelligence (AI) into the heart of biogas decision-making.

Their core idea was to build a multi-layer digital twin—a system capable of simulating and optimising the full AD process using real-world data, advanced modelling techniques, and machine learning. Unlike existing models, this hybrid approach would be capable of adapting to changes in feedstock, microbial activity, and operating conditions in real time.

The project brought together a multidisciplinary team with expertise in AI, systems microbiology, process optimisation, and life-cycle assessment. Together, they developed:

• A hybrid machine-learning digital twin of the AD process, trained on data from industry partners and new experimental findings
• A site-wide optimisation model to simulate economic and environmental performance across multiple operational layers
• Decision-support tools for feedstock procurement, gas forecasting, and emissions tracking

BiofuelAI emerged as a commercial spinout from this research, with the mission of making these tools accessible and actionable for biogas operators.
By turning cutting-edge academic research into practical software, BiofuelAI helps operators:

• Evaluate and compare incoming feedstock batches
• Forecast gas output and process stability
• Recommend optimal feedstock mixes in real-time
• Monitor environmental impacts and emissions performance
• Increase uptime, reduce cost, and improve plant flexibility

This technology promises to make biogas more competitive, reduce dependence on fossil-derived natural gas, and contribute meaningfully to the UK’s Net Zero strategy.

Although developed for AD, the digital twin and AI systems designed through the project have broader relevance. Similar microbial processes are used in wastewater treatment, fermentation, biopharmaceuticals, and food production. The methods pioneered at University of Surrey—and now being scaled by BiofuelAI—could support efficiency gains across the wider bioprocessing industry.

Supported by a national network of partners and aligned with the UK’s ambitions to lead in clean energy innovation, BiofuelAI represents a step-change in how biological energy systems are understood, managed, and optimised.

As Alan Beesley, co-founder of BiofuelAI, puts it:
“In AD, what we can’t predict, we can’t control. And what we can’t control, we can’t optimise. With BiofuelAI, we’re turning prediction into performance.”